root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 03:14:35 +00:00
Code Issues Actions 7 Packages Projects Releases Wiki Activity

ggml : sync latest ggml repo (#1924)

Browse Source 
* ggml : sync latest ggml repo

* ggml : remove unused comments

* ggml : asserts
This commit is contained in:
Georgi Gerganov 2023-06-19 18:12:33 +03:00 committed by GitHub
parent 1e3abfcef0
commit b97ca431db
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 844 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

793
ggml.c
View File

File diff suppressed because it is too large Load Diff

142
ggml.h
View File

@ -303,6 +303,7 @@ extern "C" {
GGML_OP_STEP, GGML_OP_STEP,
GGML_OP_RELU, GGML_OP_RELU,
GGML_OP_GELU, GGML_OP_GELU,
GGML_OP_GELU_QUICK,
GGML_OP_SILU, GGML_OP_SILU,
GGML_OP_SILU_BACK, GGML_OP_SILU_BACK,
GGML_OP_NORM, // normalize GGML_OP_NORM, // normalize
@ -331,12 +332,15 @@ extern "C" {
GGML_OP_ROPE_BACK, GGML_OP_ROPE_BACK,
GGML_OP_ALIBI, GGML_OP_ALIBI,
GGML_OP_CLAMP, GGML_OP_CLAMP,
GGML_OP_CONV_1D_1S, GGML_OP_CONV_1D_S1_PH,
GGML_OP_CONV_1D_2S, GGML_OP_CONV_1D_S2_PH,
GGML_OP_CONV_2D_SK_P0,
GGML_OP_FLASH_ATTN, GGML_OP_FLASH_ATTN,
GGML_OP_FLASH_FF, GGML_OP_FLASH_FF,
GGML_OP_FLASH_ATTN_BACK, GGML_OP_FLASH_ATTN_BACK,
GGML_OP_WIN_PART,
GGML_OP_WIN_UNPART,
GGML_OP_MAP_UNARY, GGML_OP_MAP_UNARY,
GGML_OP_MAP_BINARY, GGML_OP_MAP_BINARY,
@ -558,7 +562,7 @@ extern "C" {
GGML_API float * ggml_get_data_f32(const struct ggml_tensor * tensor); GGML_API float * ggml_get_data_f32(const struct ggml_tensor * tensor);
GGML_API const char * ggml_get_name(const struct ggml_tensor * tensor); GGML_API const char * ggml_get_name(const struct ggml_tensor * tensor);
GGML_API void ggml_set_name(struct ggml_tensor * tensor, const char * name); GGML_API struct ggml_tensor * ggml_set_name(struct ggml_tensor * tensor, const char * name);
// //
// operations on tensors with backpropagation // operations on tensors with backpropagation
@ -611,24 +615,47 @@ extern "C" {
struct ggml_tensor * a, struct ggml_tensor * a,
struct ggml_tensor * b); struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_sub_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_mul( GGML_API struct ggml_tensor * ggml_mul(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a, struct ggml_tensor * a,
struct ggml_tensor * b); struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_mul_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_div( GGML_API struct ggml_tensor * ggml_div(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a, struct ggml_tensor * a,
struct ggml_tensor * b); struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_div_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_sqr( GGML_API struct ggml_tensor * ggml_sqr(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_sqr_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_sqrt( GGML_API struct ggml_tensor * ggml_sqrt(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_sqrt_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_log( GGML_API struct ggml_tensor * ggml_log(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
@ -668,31 +695,67 @@ extern "C" {
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_abs_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_sgn( GGML_API struct ggml_tensor * ggml_sgn(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_sgn_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_neg( GGML_API struct ggml_tensor * ggml_neg(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_neg_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_step( GGML_API struct ggml_tensor * ggml_step(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_step_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_relu( GGML_API struct ggml_tensor * ggml_relu(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_relu_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
// TODO: double-check this computation is correct // TODO: double-check this computation is correct
GGML_API struct ggml_tensor * ggml_gelu( GGML_API struct ggml_tensor * ggml_gelu(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_gelu_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_gelu_quick(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_gelu_quick_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_silu( GGML_API struct ggml_tensor * ggml_silu(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_silu_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
// a - x // a - x
// b - dy // b - dy
GGML_API struct ggml_tensor * ggml_silu_back( GGML_API struct ggml_tensor * ggml_silu_back(
@ -706,10 +769,18 @@ extern "C" {
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_norm_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_rms_norm( GGML_API struct ggml_tensor * ggml_rms_norm(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a); struct ggml_tensor * a);
GGML_API struct ggml_tensor * ggml_rms_norm_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);
// a - x // a - x
// b - dy // b - dy
GGML_API struct ggml_tensor * ggml_rms_norm_back( GGML_API struct ggml_tensor * ggml_rms_norm_back(
@ -999,16 +1070,55 @@ extern "C" {
float min, float min,
float max); float max);
// padding = 1 // TODO: implement general-purpose convolutions
// GGML_API struct ggml_tensor * ggml_conv_1d(
// struct ggml_context * ctx,
// struct ggml_tensor * a,
// struct ggml_tensor * b,
// int s0
// int p0,
// int d0);
//
// GGML_API struct ggml_tensor * ggml_conv_2d(
// struct ggml_context * ctx,
// struct ggml_tensor * a,
// struct ggml_tensor * b,
// int s0,
// int s1,
// int p0,
// int p1,
// int d0,
// int d1);
// padding = half
// TODO: we don't support extra parameters for now // TODO: we don't support extra parameters for now
// that's why we are hard-coding the stride, padding, and dilation // that's why we are hard-coding the stride, padding, and dilation
// not great .. // not great ..
GGML_API struct ggml_tensor * ggml_conv_1d_1s( // example:
// a: 3 80 768 1
// b: 3000 80 1 1
// res: 3000 768 1 1
// used in whisper
GGML_API struct ggml_tensor * ggml_conv_1d_s1_ph(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a, struct ggml_tensor * a,
struct ggml_tensor * b); struct ggml_tensor * b);
GGML_API struct ggml_tensor * ggml_conv_1d_2s( // used in whisper
GGML_API struct ggml_tensor * ggml_conv_1d_s2_ph(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
// kernel size is a->ne[0] x a->ne[1]
// stride is equal to kernel size
// padding is zero
// example:
// a: 16 16 3 768
// b: 1024 1024 3 1
// res: 64 64 768 1
// used in sam
GGML_API struct ggml_tensor * ggml_conv_2d_sk_p0(
struct ggml_context * ctx, struct ggml_context * ctx,
struct ggml_tensor * a, struct ggml_tensor * a,
struct ggml_tensor * b); struct ggml_tensor * b);
@ -1036,6 +1146,26 @@ extern "C" {
struct ggml_tensor * c0, struct ggml_tensor * c0,
struct ggml_tensor * c1); struct ggml_tensor * c1);
// partition into non-overlapping windows with padding if needed
// example:
// a: 768 64 64 1
// w: 14
// res: 768 14 14 25
// used in sam
GGML_API struct ggml_tensor * ggml_win_part(
struct ggml_context * ctx,
struct ggml_tensor * a,
int w);
// reverse of ggml_win_part
// used in sam
GGML_API struct ggml_tensor * ggml_win_unpart(
struct ggml_context * ctx,
struct ggml_tensor * a,
int w0,
int h0,
int w);
// Mapping operations // Mapping operations
typedef void (*ggml_unary_op_f32_t)(const int, float *, const float *); typedef void (*ggml_unary_op_f32_t)(const int, float *, const float *);
typedef void (*ggml_binary_op_f32_t)(const int, float *, const float *, const float *); typedef void (*ggml_binary_op_f32_t)(const int, float *, const float *, const float *);